projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: do not set fs read-only if rec[0] is empty while committing truncate
[deliverable/linux.git] / fs / ocfs2 / super.c
CommitLineData
ccd979bd
MF		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * super.c
5 *
6 * load/unload driver, mount/dismount volumes
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#include <linux/module.h>
27#include <linux/fs.h>
28#include <linux/types.h>
29#include <linux/slab.h>
30#include <linux/highmem.h>
ccd979bd
MF		 31#include <linux/init.h>
32#include <linux/random.h>
33#include <linux/statfs.h>
34#include <linux/moduleparam.h>
35#include <linux/blkdev.h>
36#include <linux/socket.h>
37#include <linux/inet.h>
38#include <linux/parser.h>
39#include <linux/crc32.h>
40#include <linux/debugfs.h>
d550071c 41#include <linux/mount.h>
6953b4c0 42#include <linux/seq_file.h>
19ece546 43#include <linux/quotaops.h>
1cfd8bd0 44#include <linux/cleancache.h>
ccd979bd 45
80a9a84d
WW		 46#define CREATE_TRACE_POINTS
47#include "ocfs2_trace.h"
48
ccd979bd
MF		 49#include <cluster/masklog.h>
50
51#include "ocfs2.h"
52
53/* this should be the only file to include a version 1 header */
54#include "ocfs1_fs_compat.h"
55
56#include "alloc.h"
a11f7e63 57#include "aops.h"
d030cc97 58#include "blockcheck.h"
ccd979bd
MF		 59#include "dlmglue.h"
60#include "export.h"
61#include "extent_map.h"
62#include "heartbeat.h"
63#include "inode.h"
64#include "journal.h"
65#include "localalloc.h"
66#include "namei.h"
67#include "slot_map.h"
68#include "super.h"
69#include "sysfile.h"
70#include "uptodate.h"
cf1d6c76 71#include "xattr.h"
9e33d69f 72#include "quota.h"
374a263e 73#include "refcounttree.h"
b89c5428 74#include "suballoc.h"
ccd979bd
MF		 75
76#include "buffer_head_io.h"
77
1a5c4e2a 78static struct kmem_cache *ocfs2_inode_cachep;
9e33d69f
JK		 79struct kmem_cache *ocfs2_dquot_cachep;
80struct kmem_cache *ocfs2_qf_chunk_cachep;
ccd979bd 81
25985edc 82/* OCFS2 needs to schedule several different types of work which
ccd979bd
MF		 83 * require cluster locking, disk I/O, recovery waits, etc. Since these
84 * types of work tend to be heavy we avoid using the kernel events
85 * workqueue and schedule on our own. */
86struct workqueue_struct *ocfs2_wq = NULL;
87
1a5c4e2a 88static struct dentry *ocfs2_debugfs_root;
ccd979bd
MF		 89
90MODULE_AUTHOR("Oracle");
91MODULE_LICENSE("GPL");
ff8fb335 92MODULE_DESCRIPTION("OCFS2 cluster file system");
ccd979bd 93
c0123ade
TY		 94struct mount_options
95{
d147b3d6 96 unsigned long commit_interval;
c0123ade
TY		 97 unsigned long mount_opt;
98 unsigned int atime_quantum;
99 signed short slot;
73c8a800 100 int localalloc_opt;
d02f00cc 101 unsigned int resv_level;
83f92318 102 int dir_resv_level;
b61817e1 103 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
c0123ade
TY		 104};
105
ccd979bd 106static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 107 struct mount_options *mopt,
baf4661a 108 int is_remount);
5297aad8
JK		 109static int ocfs2_check_set_options(struct super_block *sb,
110 struct mount_options *options);
34c80b1d 111static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
ccd979bd
MF		 112static void ocfs2_put_super(struct super_block *sb);
113static int ocfs2_mount_volume(struct super_block *sb);
114static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
115static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
116static int ocfs2_initialize_mem_caches(void);
117static void ocfs2_free_mem_caches(void);
118static void ocfs2_delete_osb(struct ocfs2_super *osb);
119
726c3342 120static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 121
122static int ocfs2_sync_fs(struct super_block *sb, int wait);
123
124static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
125static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 126static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 127static int ocfs2_check_volume(struct ocfs2_super *osb);
128static int ocfs2_verify_volume(struct ocfs2_dinode *di,
129 struct buffer_head *bh,
73be192b
JB		 130 u32 sectsize,
131 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 132static int ocfs2_initialize_super(struct super_block *sb,
133 struct buffer_head *bh,
73be192b
JB		 134 int sector_size,
135 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 136static int ocfs2_get_sector(struct super_block *sb,
137 struct buffer_head **bh,
138 int block,
139 int sect_size);
ccd979bd
MF		 140static struct inode *ocfs2_alloc_inode(struct super_block *sb);
141static void ocfs2_destroy_inode(struct inode *inode);
19ece546
JK		 142static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
143static int ocfs2_enable_quotas(struct ocfs2_super *osb);
144static void ocfs2_disable_quotas(struct ocfs2_super *osb);
ccd979bd 145
1c92ec67
JK		 146static struct dquot **ocfs2_get_dquots(struct inode *inode)
147{
148 return OCFS2_I(inode)->i_dquot;
149}
150
ee9b6d61 151static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 152 .statfs = ocfs2_statfs,
153 .alloc_inode = ocfs2_alloc_inode,
154 .destroy_inode = ocfs2_destroy_inode,
155 .drop_inode = ocfs2_drop_inode,
066d92dc 156 .evict_inode = ocfs2_evict_inode,
ccd979bd 157 .sync_fs = ocfs2_sync_fs,
ccd979bd
MF		 158 .put_super = ocfs2_put_super,
159 .remount_fs = ocfs2_remount,
d550071c 160 .show_options = ocfs2_show_options,
9e33d69f
JK		 161 .quota_read = ocfs2_quota_read,
162 .quota_write = ocfs2_quota_write,
1c92ec67 163 .get_dquots = ocfs2_get_dquots,
ccd979bd
MF		 164};
165
166enum {
167 Opt_barrier,
168 Opt_err_panic,
169 Opt_err_ro,
170 Opt_intr,
171 Opt_nointr,
172 Opt_hb_none,
173 Opt_hb_local,
2c442719 174 Opt_hb_global,
ccd979bd
MF		 175 Opt_data_ordered,
176 Opt_data_writeback,
7f1a37e3 177 Opt_atime_quantum,
baf4661a 178 Opt_slot,
d147b3d6 179 Opt_commit,
2fbe8d1e 180 Opt_localalloc,
53fc622b 181 Opt_localflocks,
b61817e1 182 Opt_stack,
cf1d6c76
TY		 183 Opt_user_xattr,
184 Opt_nouser_xattr,
12462f1d 185 Opt_inode64,
a68979b8
TY		 186 Opt_acl,
187 Opt_noacl,
19ece546
JK		 188 Opt_usrquota,
189 Opt_grpquota,
7bdb0d18
TY		 190 Opt_coherency_buffered,
191 Opt_coherency_full,
d02f00cc 192 Opt_resv_level,
83f92318 193 Opt_dir_resv_level,
1dfeb768 194 Opt_journal_async_commit,
7d0fb914 195 Opt_err_cont,
ccd979bd
MF		 196 Opt_err,
197};
198
a447c093 199static const match_table_t tokens = {
ccd979bd
MF		 200 {Opt_barrier, "barrier=%u"},
201 {Opt_err_panic, "errors=panic"},
202 {Opt_err_ro, "errors=remount-ro"},
203 {Opt_intr, "intr"},
204 {Opt_nointr, "nointr"},
205 {Opt_hb_none, OCFS2_HB_NONE},
206 {Opt_hb_local, OCFS2_HB_LOCAL},
2c442719 207 {Opt_hb_global, OCFS2_HB_GLOBAL},
ccd979bd
MF		 208 {Opt_data_ordered, "data=ordered"},
209 {Opt_data_writeback, "data=writeback"},
7f1a37e3 210 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 211 {Opt_slot, "preferred_slot=%u"},
d147b3d6 212 {Opt_commit, "commit=%u"},
2fbe8d1e 213 {Opt_localalloc, "localalloc=%d"},
53fc622b 214 {Opt_localflocks, "localflocks"},
b61817e1 215 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 216 {Opt_user_xattr, "user_xattr"},
217 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 218 {Opt_inode64, "inode64"},
a68979b8
TY		 219 {Opt_acl, "acl"},
220 {Opt_noacl, "noacl"},
19ece546
JK		 221 {Opt_usrquota, "usrquota"},
222 {Opt_grpquota, "grpquota"},
7bdb0d18
TY		 223 {Opt_coherency_buffered, "coherency=buffered"},
224 {Opt_coherency_full, "coherency=full"},
d02f00cc 225 {Opt_resv_level, "resv_level=%u"},
83f92318 226 {Opt_dir_resv_level, "dir_resv_level=%u"},
1dfeb768 227 {Opt_journal_async_commit, "journal_async_commit"},
7d0fb914 228 {Opt_err_cont, "errors=continue"},
ccd979bd
MF		 229 {Opt_err, NULL}
230};
231
50397507
SM		 232#ifdef CONFIG_DEBUG_FS
233static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
234{
50397507
SM		 235 struct ocfs2_cluster_connection *cconn = osb->cconn;
236 struct ocfs2_recovery_map *rm = osb->recovery_map;
df152c24
SM		 237 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
238 int i, out = 0;
50397507
SM		 239
240 out += snprintf(buf + out, len - out,
241 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
242 "Device", osb->dev_str, osb->uuid_str,
243 osb->fs_generation, osb->vol_label);
244
245 out += snprintf(buf + out, len - out,
246 "%10s => State: %d Flags: 0x%lX\n", "Volume",
247 atomic_read(&osb->vol_state), osb->osb_flags);
248
249 out += snprintf(buf + out, len - out,
250 "%10s => Block: %lu Cluster: %d\n", "Sizes",
251 osb->sb->s_blocksize, osb->s_clustersize);
252
253 out += snprintf(buf + out, len - out,
254 "%10s => Compat: 0x%X Incompat: 0x%X "
255 "ROcompat: 0x%X\n",
256 "Features", osb->s_feature_compat,
257 osb->s_feature_incompat, osb->s_feature_ro_compat);
258
259 out += snprintf(buf + out, len - out,
260 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
261 osb->s_mount_opt, osb->s_atime_quantum);
262
c3d38840
SM		 263 if (cconn) {
264 out += snprintf(buf + out, len - out,
265 "%10s => Stack: %s Name: %*s "
266 "Version: %d.%d\n", "Cluster",
267 (*osb->osb_cluster_stack == '\0' ?
268 "o2cb" : osb->osb_cluster_stack),
269 cconn->cc_namelen, cconn->cc_name,
270 cconn->cc_version.pv_major,
271 cconn->cc_version.pv_minor);
272 }
50397507
SM		 273
274 spin_lock(&osb->dc_task_lock);
275 out += snprintf(buf + out, len - out,
276 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
277 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 278 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
279 osb->blocked_lock_count, osb->dc_wake_sequence,
280 osb->dc_work_sequence);
50397507
SM		 281 spin_unlock(&osb->dc_task_lock);
282
283 spin_lock(&osb->osb_lock);
284 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
285 "Recovery",
286 (osb->recovery_thread_task ?
287 task_pid_nr(osb->recovery_thread_task) : -1));
288 if (rm->rm_used == 0)
289 out += snprintf(buf + out, len - out, " None\n");
290 else {
291 for (i = 0; i < rm->rm_used; i++)
292 out += snprintf(buf + out, len - out, " %d",
293 rm->rm_entries[i]);
294 out += snprintf(buf + out, len - out, "\n");
295 }
296 spin_unlock(&osb->osb_lock);
297
298 out += snprintf(buf + out, len - out,
8fa9d17f 299 "%10s => Pid: %d Interval: %lu\n", "Commit",
c3d38840 300 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
8fa9d17f 301 osb->osb_commit_interval);
50397507
SM		 302
303 out += snprintf(buf + out, len - out,
c3d38840 304 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 305 "Journal", osb->journal->j_state,
c3d38840
SM		 306 osb->journal->j_trans_id,
307 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 308
309 out += snprintf(buf + out, len - out,
310 "%10s => GlobalAllocs: %d LocalAllocs: %d "
311 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
312 "Stats",
313 atomic_read(&osb->alloc_stats.bitmap_data),
314 atomic_read(&osb->alloc_stats.local_data),
315 atomic_read(&osb->alloc_stats.bg_allocs),
316 atomic_read(&osb->alloc_stats.moves),
317 atomic_read(&osb->alloc_stats.bg_extends));
318
319 out += snprintf(buf + out, len - out,
320 "%10s => State: %u Descriptor: %llu Size: %u bits "
321 "Default: %u bits\n",
322 "LocalAlloc", osb->local_alloc_state,
323 (unsigned long long)osb->la_last_gd,
324 osb->local_alloc_bits, osb->local_alloc_default_bits);
325
326 spin_lock(&osb->osb_lock);
327 out += snprintf(buf + out, len - out,
b89c5428
TY		 328 "%10s => InodeSlot: %d StolenInodes: %d, "
329 "MetaSlot: %d StolenMeta: %d\n", "Steal",
50397507 330 osb->s_inode_steal_slot,
b89c5428
TY		 331 atomic_read(&osb->s_num_inodes_stolen),
332 osb->s_meta_steal_slot,
333 atomic_read(&osb->s_num_meta_stolen));
50397507
SM		 334 spin_unlock(&osb->osb_lock);
335
df152c24
SM		 336 out += snprintf(buf + out, len - out, "OrphanScan => ");
337 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
338 os->os_count, os->os_seqno);
339 out += snprintf(buf + out, len - out, " Last Scan: ");
340 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
341 out += snprintf(buf + out, len - out, "Disabled\n");
342 else
343 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
344 (get_seconds() - os->os_scantime.tv_sec));
345
50397507
SM		 346 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
347 "Slots", "Num", "RecoGen");
50397507
SM		 348 for (i = 0; i < osb->max_slots; ++i) {
349 out += snprintf(buf + out, len - out,
350 "%10s %c %3d %10d\n",
351 " ",
352 (i == osb->slot_num ? '*' : ' '),
353 i, osb->slot_recovery_generations[i]);
354 }
355
356 return out;
357}
358
359static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
360{
361 struct ocfs2_super *osb = inode->i_private;
362 char *buf = NULL;
363
364 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
365 if (!buf)
366 goto bail;
367
368 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
369
370 file->private_data = buf;
371
372 return 0;
373bail:
374 return -ENOMEM;
375}
376
377static int ocfs2_debug_release(struct inode *inode, struct file *file)
378{
379 kfree(file->private_data);
380 return 0;
381}
382
383static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
384 size_t nbytes, loff_t *ppos)
385{
386 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
387 i_size_read(file->f_mapping->host));
388}
389#else
390static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
391{
392 return 0;
393}
394static int ocfs2_debug_release(struct inode *inode, struct file *file)
395{
396 return 0;
397}
398static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
399 size_t nbytes, loff_t *ppos)
400{
401 return 0;
402}
403#endif /* CONFIG_DEBUG_FS */
404
828c0950 405static const struct file_operations ocfs2_osb_debug_fops = {
50397507
SM		 406 .open = ocfs2_osb_debug_open,
407 .release = ocfs2_debug_release,
408 .read = ocfs2_debug_read,
409 .llseek = generic_file_llseek,
410};
411
ccd979bd
MF		 412static int ocfs2_sync_fs(struct super_block *sb, int wait)
413{
bddb8eb3 414 int status;
ccd979bd
MF		 415 tid_t target;
416 struct ocfs2_super *osb = OCFS2_SB(sb);
417
ccd979bd
MF		 418 if (ocfs2_is_hard_readonly(osb))
419 return -EROFS;
420
421 if (wait) {
422 status = ocfs2_flush_truncate_log(osb);
423 if (status < 0)
424 mlog_errno(status);
425 } else {
426 ocfs2_schedule_truncate_log_flush(osb, 0);
427 }
428
2b4e30fb
JB		 429 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
430 &target)) {
ccd979bd 431 if (wait)
2b4e30fb
JB		 432 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
433 target);
ccd979bd
MF		 434 }
435 return 0;
436}
437
1a224ad1
JK		 438static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
439{
440 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
441 && (ino == USER_QUOTA_SYSTEM_INODE
442 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
443 return 0;
444 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
445 && (ino == GROUP_QUOTA_SYSTEM_INODE
446 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
447 return 0;
448 return 1;
449}
450
ccd979bd
MF		 451static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
452{
453 struct inode *new = NULL;
454 int status = 0;
455 int i;
456
5fa0613e 457 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 458 if (IS_ERR(new)) {
459 status = PTR_ERR(new);
460 mlog_errno(status);
461 goto bail;
462 }
463 osb->root_inode = new;
464
5fa0613e 465 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 466 if (IS_ERR(new)) {
467 status = PTR_ERR(new);
468 mlog_errno(status);
469 goto bail;
470 }
471 osb->sys_root_inode = new;
472
473 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
474 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 475 if (!ocfs2_need_system_inode(osb, i))
476 continue;
ccd979bd
MF		 477 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
478 if (!new) {
479 ocfs2_release_system_inodes(osb);
480 status = -EINVAL;
481 mlog_errno(status);
482 /* FIXME: Should ERROR_RO_FS */
483 mlog(ML_ERROR, "Unable to load system inode %d, "
484 "possibly corrupt fs?", i);
485 goto bail;
486 }
487 // the array now has one ref, so drop this one
488 iput(new);
489 }
490
491bail:
c1e8d35e
TM		 492 if (status)
493 mlog_errno(status);
ccd979bd
MF		 494 return status;
495}
496
497static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
498{
499 struct inode *new = NULL;
500 int status = 0;
501 int i;
502
ccd979bd
MF		 503 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
504 i < NUM_SYSTEM_INODES;
505 i++) {
1a224ad1
JK		 506 if (!ocfs2_need_system_inode(osb, i))
507 continue;
ccd979bd
MF		 508 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
509 if (!new) {
510 ocfs2_release_system_inodes(osb);
511 status = -EINVAL;
512 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
513 status, i, osb->slot_num);
514 goto bail;
515 }
516 /* the array now has one ref, so drop this one */
517 iput(new);
518 }
519
520bail:
c1e8d35e
TM		 521 if (status)
522 mlog_errno(status);
ccd979bd
MF		 523 return status;
524}
525
bddb8eb3 526static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 527{
bddb8eb3 528 int i;
ccd979bd
MF		 529 struct inode *inode;
530
b4d693fc
TM		 531 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
532 inode = osb->global_system_inodes[i];
ccd979bd
MF		 533 if (inode) {
534 iput(inode);
b4d693fc 535 osb->global_system_inodes[i] = NULL;
ccd979bd
MF		 536 }
537 }
538
539 inode = osb->sys_root_inode;
540 if (inode) {
541 iput(inode);
542 osb->sys_root_inode = NULL;
543 }
544
545 inode = osb->root_inode;
546 if (inode) {
547 iput(inode);
548 osb->root_inode = NULL;
549 }
550
b4d693fc 551 if (!osb->local_system_inodes)
c1e8d35e 552 return;
b4d693fc
TM		 553
554 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
555 if (osb->local_system_inodes[i]) {
556 iput(osb->local_system_inodes[i]);
557 osb->local_system_inodes[i] = NULL;
558 }
559 }
560
561 kfree(osb->local_system_inodes);
562 osb->local_system_inodes = NULL;
ccd979bd
MF		 563}
564
565/* We're allocating fs objects, use GFP_NOFS */
566static struct inode *ocfs2_alloc_inode(struct super_block *sb)
567{
568 struct ocfs2_inode_info *oi;
569
e6b4f8da 570 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 571 if (!oi)
572 return NULL;
573
2931cdcb
DW		 574 oi->i_sync_tid = 0;
575 oi->i_datasync_tid = 0;
1c92ec67 576 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
2931cdcb 577
2b4e30fb 578 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
ccd979bd
MF		 579 return &oi->vfs_inode;
580}
581
fa0d7e3d 582static void ocfs2_i_callback(struct rcu_head *head)
ccd979bd 583{
fa0d7e3d 584 struct inode *inode = container_of(head, struct inode, i_rcu);
ccd979bd
MF		 585 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
586}
587
fa0d7e3d
NP		 588static void ocfs2_destroy_inode(struct inode *inode)
589{
590 call_rcu(&inode->i_rcu, ocfs2_i_callback);
591}
592
5a254031
MF		 593static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
594 unsigned int cbits)
ccd979bd 595{
5a254031
MF		 596 unsigned int bytes = 1 << cbits;
597 unsigned int trim = bytes;
598 unsigned int bitshift = 32;
599
600 /*
601 * i_size and all block offsets in ocfs2 are always 64 bits
602 * wide. i_clusters is 32 bits, in cluster-sized units. So on
603 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 604 */
605
606#if BITS_PER_LONG == 32
90c699a9 607# if defined(CONFIG_LBDAF)
2ecd05ae 608 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 609 /*
610 * We might be limited by page cache size.
611 */
612 if (bytes > PAGE_CACHE_SIZE) {
613 bytes = PAGE_CACHE_SIZE;
614 trim = 1;
615 /*
616 * Shift by 31 here so that we don't get larger than
617 * MAX_LFS_FILESIZE
618 */
619 bitshift = 31;
620 }
ccd979bd 621# else
5a254031
MF		 622 /*
623 * We are limited by the size of sector_t. Use block size, as
624 * that's what we expose to the VFS.
625 */
626 bytes = 1 << bbits;
627 trim = 1;
628 bitshift = 31;
ccd979bd
MF		 629# endif
630#endif
631
5a254031
MF		 632 /*
633 * Trim by a whole cluster when we can actually approach the
634 * on-disk limits. Otherwise we can overflow i_clusters when
635 * an extent start is at the max offset.
636 */
637 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 638}
639
640static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
641{
642 int incompat_features;
643 int ret = 0;
c0123ade 644 struct mount_options parsed_options;
ccd979bd 645 struct ocfs2_super *osb = OCFS2_SB(sb);
2c442719 646 u32 tmp;
ccd979bd 647
02b9984d
TT		 648 sync_filesystem(sb);
649
5297aad8
JK		 650 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
651 !ocfs2_check_set_options(sb, &parsed_options)) {
ccd979bd
MF		 652 ret = -EINVAL;
653 goto out;
654 }
655
2c442719
SM		 656 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
657 OCFS2_MOUNT_HB_NONE;
658 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
ccd979bd
MF		 659 ret = -EINVAL;
660 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
661 goto out;
662 }
663
664 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 665 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 666 ret = -EINVAL;
667 mlog(ML_ERROR, "Cannot change data mode on remount\n");
668 goto out;
669 }
670
12462f1d
JB		 671 /* Probably don't want this on remount; it might
672 * mess with other nodes */
673 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
674 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
675 ret = -EINVAL;
676 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
677 goto out;
678 }
679
ccd979bd
MF		 680 /* We're going to/from readonly mode. */
681 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK		 682 /* Disable quota accounting before remounting RO */
683 if (*flags & MS_RDONLY) {
684 ret = ocfs2_susp_quotas(osb, 0);
685 if (ret < 0)
686 goto out;
687 }
ccd979bd
MF		 688 /* Lock here so the check of HARD_RO and the potential
689 * setting of SOFT_RO is atomic. */
690 spin_lock(&osb->osb_lock);
691 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
692 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
693 ret = -EROFS;
694 goto unlock_osb;
695 }
696
697 if (*flags & MS_RDONLY) {
ccd979bd
MF		 698 sb->s_flags |= MS_RDONLY;
699 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
700 } else {
ccd979bd
MF		 701 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
702 mlog(ML_ERROR, "Cannot remount RDWR "
703 "filesystem due to previous errors.\n");
704 ret = -EROFS;
705 goto unlock_osb;
706 }
707 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
708 if (incompat_features) {
709 mlog(ML_ERROR, "Cannot remount RDWR because "
710 "of unsupported optional features "
711 "(%x).\n", incompat_features);
712 ret = -EINVAL;
713 goto unlock_osb;
714 }
715 sb->s_flags &= ~MS_RDONLY;
716 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
717 }
32a42d39 718 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
ccd979bd
MF		 719unlock_osb:
720 spin_unlock(&osb->osb_lock);
19ece546
JK		 721 /* Enable quota accounting after remounting RW */
722 if (!ret && !(*flags & MS_RDONLY)) {
723 if (sb_any_quota_suspended(sb))
724 ret = ocfs2_susp_quotas(osb, 1);
725 else
726 ret = ocfs2_enable_quotas(osb);
727 if (ret < 0) {
728 /* Return back changes... */
729 spin_lock(&osb->osb_lock);
730 sb->s_flags |= MS_RDONLY;
731 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
732 spin_unlock(&osb->osb_lock);
733 goto out;
734 }
735 }
ccd979bd
MF		 736 }
737
738 if (!ret) {
ccd979bd
MF		 739 /* Only save off the new mount options in case of a successful
740 * remount. */
c0123ade
TY		 741 osb->s_mount_opt = parsed_options.mount_opt;
742 osb->s_atime_quantum = parsed_options.atime_quantum;
743 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 744 if (parsed_options.commit_interval)
745 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 746
747 if (!ocfs2_is_hard_readonly(osb))
748 ocfs2_set_journal_params(osb);
57b09bb5
JK		 749
750 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
751 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
752 MS_POSIXACL : 0);
ccd979bd
MF		 753 }
754out:
755 return ret;
756}
757
758static int ocfs2_sb_probe(struct super_block *sb,
759 struct buffer_head **bh,
73be192b
JB		 760 int *sector_size,
761 struct ocfs2_blockcheck_stats *stats)
ccd979bd 762{
bddb8eb3 763 int status, tmpstat;
ccd979bd
MF		 764 struct ocfs1_vol_disk_hdr *hdr;
765 struct ocfs2_dinode *di;
766 int blksize;
767
768 *bh = NULL;
769
770 /* may be > 512 */
e1defc4f 771 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF		 772 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
773 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
774 *sector_size, OCFS2_MAX_BLOCKSIZE);
775 status = -EINVAL;
776 goto bail;
777 }
778
779 /* Can this really happen? */
780 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
781 *sector_size = OCFS2_MIN_BLOCKSIZE;
782
783 /* check block zero for old format */
784 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
785 if (status < 0) {
786 mlog_errno(status);
787 goto bail;
788 }
789 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
790 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
791 mlog(ML_ERROR, "incompatible version: %u.%u\n",
792 hdr->major_version, hdr->minor_version);
793 status = -EINVAL;
794 }
795 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
796 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
797 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
798 hdr->signature);
799 status = -EINVAL;
800 }
801 brelse(*bh);
802 *bh = NULL;
803 if (status < 0) {
804 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
805 "upgraded before mounting with ocfs v2\n");
806 goto bail;
807 }
808
809 /*
810 * Now check at magic offset for 512, 1024, 2048, 4096
811 * blocksizes. 4096 is the maximum blocksize because it is
812 * the minimum clustersize.
813 */
814 status = -EINVAL;
815 for (blksize = *sector_size;
816 blksize <= OCFS2_MAX_BLOCKSIZE;
817 blksize <<= 1) {
818 tmpstat = ocfs2_get_sector(sb, bh,
819 OCFS2_SUPER_BLOCK_BLKNO,
820 blksize);
821 if (tmpstat < 0) {
822 status = tmpstat;
823 mlog_errno(status);
fb5cbe9e 824 break;
ccd979bd
MF		 825 }
826 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b 827 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
1c1d9793 828 spin_lock_init(&stats->b_lock);
fb5cbe9e
JB		 829 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
830 if (tmpstat < 0) {
831 brelse(*bh);
832 *bh = NULL;
833 }
834 if (tmpstat != -EAGAIN) {
835 status = tmpstat;
ccd979bd 836 break;
fb5cbe9e 837 }
ccd979bd
MF		 838 }
839
840bail:
841 return status;
842}
843
c271c5c2
SM		 844static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
845{
2c442719
SM		 846 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
847
848 if (osb->s_mount_opt & hb_enabled) {
849 if (ocfs2_mount_local(osb)) {
c271c5c2
SM		 850 mlog(ML_ERROR, "Cannot heartbeat on a locally "
851 "mounted device.\n");
852 return -EINVAL;
853 }
2c442719 854 if (ocfs2_userspace_stack(osb)) {
b61817e1
JB		 855 mlog(ML_ERROR, "Userspace stack expected, but "
856 "o2cb heartbeat arguments passed to mount\n");
857 return -EINVAL;
858 }
2c442719
SM		 859 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
860 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
861 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
862 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
863 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
864 return -EINVAL;
865 }
b61817e1
JB		 866 }
867
2c442719 868 if (!(osb->s_mount_opt & hb_enabled)) {
b61817e1
JB		 869 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
870 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM		 871 mlog(ML_ERROR, "Heartbeat has to be started to mount "
872 "a read-write clustered device.\n");
873 return -EINVAL;
874 }
875 }
876
877 return 0;
878}
879
b61817e1
JB		 880/*
881 * If we're using a userspace stack, mount should have passed
882 * a name that matches the disk. If not, mount should not
883 * have passed a stack.
884 */
885static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
886 struct mount_options *mopt)
887{
888 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
889 mlog(ML_ERROR,
890 "cluster stack passed to mount, but this filesystem "
891 "does not support it\n");
892 return -EINVAL;
893 }
894
895 if (ocfs2_userspace_stack(osb) &&
896 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
897 OCFS2_STACK_LABEL_LEN)) {
898 mlog(ML_ERROR,
899 "cluster stack passed to mount (\"%s\") does not "
900 "match the filesystem (\"%s\")\n",
901 mopt->cluster_stack,
902 osb->osb_cluster_stack);
903 return -EINVAL;
904 }
905
906 return 0;
907}
908
19ece546
JK		 909static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
910{
911 int type;
912 struct super_block *sb = osb->sb;
52362810
JK		 913 unsigned int feature[OCFS2_MAXQUOTAS] = {
914 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
915 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
19ece546
JK		 916 int status = 0;
917
52362810 918 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 919 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
920 continue;
921 if (unsuspend)
0f0dd62f 922 status = dquot_resume(sb, type);
eea7feb0
JK		 923 else {
924 struct ocfs2_mem_dqinfo *oinfo;
925
926 /* Cancel periodic syncing before suspending */
927 oinfo = sb_dqinfo(sb, type)->dqi_priv;
928 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
0f0dd62f 929 status = dquot_suspend(sb, type);
eea7feb0 930 }
19ece546
JK		 931 if (status < 0)
932 break;
933 }
934 if (status < 0)
935 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
936 "remount (error = %d).\n", status);
937 return status;
938}
939
940static int ocfs2_enable_quotas(struct ocfs2_super *osb)
941{
52362810 942 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
19ece546 943 struct super_block *sb = osb->sb;
52362810
JK		 944 unsigned int feature[OCFS2_MAXQUOTAS] = {
945 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
946 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
947 unsigned int ino[OCFS2_MAXQUOTAS] = {
948 LOCAL_USER_QUOTA_SYSTEM_INODE,
19ece546
JK		 949 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
950 int status;
951 int type;
952
953 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
52362810 954 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 955 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
956 continue;
957 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
958 osb->slot_num);
959 if (!inode[type]) {
960 status = -ENOENT;
961 goto out_quota_off;
962 }
287a8095
CH		 963 status = dquot_enable(inode[type], type, QFMT_OCFS2,
964 DQUOT_USAGE_ENABLED);
19ece546
JK		 965 if (status < 0)
966 goto out_quota_off;
967 }
968
52362810 969 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK		 970 iput(inode[type]);
971 return 0;
972out_quota_off:
973 ocfs2_disable_quotas(osb);
52362810 974 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK		 975 iput(inode[type]);
976 mlog_errno(status);
977 return status;
978}
979
980static void ocfs2_disable_quotas(struct ocfs2_super *osb)
981{
982 int type;
983 struct inode *inode;
984 struct super_block *sb = osb->sb;
c06bcbfa 985 struct ocfs2_mem_dqinfo *oinfo;
19ece546
JK		 986
987 /* We mostly ignore errors in this function because there's not much
988 * we can do when we see them */
52362810 989 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 990 if (!sb_has_quota_loaded(sb, type))
991 continue;
c06bcbfa
JK		 992 /* Cancel periodic syncing before we grab dqonoff_mutex */
993 oinfo = sb_dqinfo(sb, type)->dqi_priv;
994 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
19ece546
JK		 995 inode = igrab(sb->s_dquot.files[type]);
996 /* Turn off quotas. This will remove all dquot structures from
997 * memory and so they will be automatically synced to global
998 * quota files */
0f0dd62f
CH		 999 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
1000 DQUOT_LIMITS_ENABLED);
19ece546
JK		 1001 if (!inode)
1002 continue;
1003 iput(inode);
1004 }
1005}
1006
ccd979bd
MF		 1007static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1008{
1009 struct dentry *root;
1010 int status, sector_size;
c0123ade 1011 struct mount_options parsed_options;
ccd979bd
MF		 1012 struct inode *inode = NULL;
1013 struct ocfs2_super *osb = NULL;
1014 struct buffer_head *bh = NULL;
49fa8140 1015 char nodestr[12];
73be192b 1016 struct ocfs2_blockcheck_stats stats;
ccd979bd 1017
32a42d39 1018 trace_ocfs2_fill_super(sb, data, silent);
ccd979bd 1019
c0123ade 1020 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 1021 status = -EINVAL;
1022 goto read_super_error;
1023 }
1024
1025 /* probe for superblock */
73be192b 1026 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 1027 if (status < 0) {
1028 mlog(ML_ERROR, "superblock probe failed!\n");
1029 goto read_super_error;
1030 }
1031
73be192b 1032 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1033 osb = OCFS2_SB(sb);
1034 if (status < 0) {
1035 mlog_errno(status);
1036 goto read_super_error;
1037 }
1038 brelse(bh);
1039 bh = NULL;
a68979b8 1040
5297aad8
JK		 1041 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1042 status = -EINVAL;
1043 goto read_super_error;
1044 }
c0123ade
TY		 1045 osb->s_mount_opt = parsed_options.mount_opt;
1046 osb->s_atime_quantum = parsed_options.atime_quantum;
1047 osb->preferred_slot = parsed_options.slot;
d147b3d6 1048 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF		 1049
1050 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1051 osb->osb_resv_level = parsed_options.resv_level;
83f92318
MF		 1052 osb->osb_dir_resv_level = parsed_options.resv_level;
1053 if (parsed_options.dir_resv_level == -1)
1054 osb->osb_dir_resv_level = parsed_options.resv_level;
1055 else
1056 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
ccd979bd 1057
b61817e1
JB		 1058 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1059 if (status)
1060 goto read_super_error;
1061
ccd979bd
MF		 1062 sb->s_magic = OCFS2_SUPER_MAGIC;
1063
9e1f1de0 1064 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
a68979b8
TY		 1065 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1066
ccd979bd
MF		 1067 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1068 * heartbeat=none */
1069 if (bdev_read_only(sb->s_bdev)) {
1070 if (!(sb->s_flags & MS_RDONLY)) {
1071 status = -EACCES;
1072 mlog(ML_ERROR, "Readonly device detected but readonly "
1073 "mount was not specified.\n");
1074 goto read_super_error;
1075 }
1076
1077 /* You should not be able to start a local heartbeat
1078 * on a readonly device. */
1079 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1080 status = -EROFS;
1081 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1082 "device.\n");
1083 goto read_super_error;
1084 }
1085
1086 status = ocfs2_check_journals_nolocks(osb);
1087 if (status < 0) {
1088 if (status == -EROFS)
1089 mlog(ML_ERROR, "Recovery required on readonly "
1090 "file system, but write access is "
1091 "unavailable.\n");
1092 else
2bd63216 1093 mlog_errno(status);
ccd979bd
MF		 1094 goto read_super_error;
1095 }
1096
1097 ocfs2_set_ro_flag(osb, 1);
1098
619c200d
SM		 1099 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1100 "Cluster services will not be used for this mount. "
1101 "Recovery will be skipped.\n", osb->dev_str);
ccd979bd
MF		 1102 }
1103
1104 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 1105 if (sb->s_flags & MS_RDONLY)
1106 ocfs2_set_ro_flag(osb, 0);
1107 }
1108
c271c5c2
SM		 1109 status = ocfs2_verify_heartbeat(osb);
1110 if (status < 0) {
1111 mlog_errno(status);
1112 goto read_super_error;
1113 }
1114
ccd979bd
MF		 1115 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1116 ocfs2_debugfs_root);
8f443e23 1117 if (!osb->osb_debug_root) {
ccd979bd
MF		 1118 status = -EINVAL;
1119 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1120 goto read_super_error;
1121 }
1122
50397507
SM		 1123 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1124 osb->osb_debug_root,
1125 osb,
1126 &ocfs2_osb_debug_fops);
8f443e23 1127 if (!osb->osb_ctxt) {
50397507
SM		 1128 status = -EINVAL;
1129 mlog_errno(status);
1130 goto read_super_error;
1131 }
1132
73be192b
JB		 1133 if (ocfs2_meta_ecc(osb)) {
1134 status = ocfs2_blockcheck_stats_debugfs_install(
1135 &osb->osb_ecc_stats,
1136 osb->osb_debug_root);
1137 if (status) {
1138 mlog(ML_ERROR,
1139 "Unable to create blockcheck statistics "
1140 "files\n");
1141 goto read_super_error;
1142 }
1143 }
1144
ccd979bd 1145 status = ocfs2_mount_volume(sb);
ccd979bd
MF		 1146 if (status < 0)
1147 goto read_super_error;
1148
be0d93f0
AV		 1149 if (osb->root_inode)
1150 inode = igrab(osb->root_inode);
1151
ccd979bd
MF		 1152 if (!inode) {
1153 status = -EIO;
1154 mlog_errno(status);
1155 goto read_super_error;
1156 }
1157
48fde701 1158 root = d_make_root(inode);
ccd979bd
MF		 1159 if (!root) {
1160 status = -ENOMEM;
1161 mlog_errno(status);
1162 goto read_super_error;
1163 }
1164
1165 sb->s_root = root;
1166
1167 ocfs2_complete_mount_recovery(osb);
1168
c271c5c2
SM		 1169 if (ocfs2_mount_local(osb))
1170 snprintf(nodestr, sizeof(nodestr), "local");
1171 else
19fdb624 1172 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1173
1174 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1175 "with %s data mode.\n",
c271c5c2 1176 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1177 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1178 "ordered");
1179
1180 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1181 wake_up(&osb->osb_mount_event);
1182
19ece546
JK		 1183 /* Now we can initialize quotas because we can afford to wait
1184 * for cluster locks recovery now. That also means that truncation
1185 * log recovery can happen but that waits for proper quota setup */
1186 if (!(sb->s_flags & MS_RDONLY)) {
1187 status = ocfs2_enable_quotas(osb);
1188 if (status < 0) {
1189 /* We have to err-out specially here because
1190 * s_root is already set */
1191 mlog_errno(status);
1192 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1193 wake_up(&osb->osb_mount_event);
19ece546
JK		 1194 return status;
1195 }
1196 }
1197
1198 ocfs2_complete_quota_recovery(osb);
1199
1200 /* Now we wake up again for processes waiting for quotas */
1201 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1202 wake_up(&osb->osb_mount_event);
1203
df152c24 1204 /* Start this when the mount is almost sure of being successful */
8b712cd5 1205 ocfs2_orphan_scan_start(osb);
df152c24 1206
ccd979bd
MF		 1207 return status;
1208
1209read_super_error:
a81cb88b 1210 brelse(bh);
ccd979bd 1211
ccd979bd
MF		 1212 if (osb) {
1213 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1214 wake_up(&osb->osb_mount_event);
1215 ocfs2_dismount_volume(sb, 1);
1216 }
1217
c1e8d35e
TM		 1218 if (status)
1219 mlog_errno(status);
ccd979bd
MF		 1220 return status;
1221}
1222
152a0836 1223static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
454e2398
DH		 1224 int flags,
1225 const char *dev_name,
152a0836 1226 void *data)
ccd979bd 1227{
152a0836 1228 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
ccd979bd
MF		 1229}
1230
1231static struct file_system_type ocfs2_fs_type = {
1232 .owner = THIS_MODULE,
1233 .name = "ocfs2",
152a0836 1234 .mount = ocfs2_mount,
8ed6b237 1235 .kill_sb = kill_block_super,
1ba9da2f 1236 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1237 .next = NULL
1238};
91417705 1239MODULE_ALIAS_FS("ocfs2");
ccd979bd 1240
5297aad8
JK		 1241static int ocfs2_check_set_options(struct super_block *sb,
1242 struct mount_options *options)
1243{
1244 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1245 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1246 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1247 mlog(ML_ERROR, "User quotas were requested, but this "
1248 "filesystem does not have the feature enabled.\n");
1249 return 0;
1250 }
1251 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1252 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1253 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1254 mlog(ML_ERROR, "Group quotas were requested, but this "
1255 "filesystem does not have the feature enabled.\n");
1256 return 0;
1257 }
1258 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1259 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1260 mlog(ML_ERROR, "ACL support requested but extended attributes "
1261 "feature is not enabled\n");
1262 return 0;
1263 }
1264 /* No ACL setting specified? Use XATTR feature... */
1265 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1266 OCFS2_MOUNT_NO_POSIX_ACL))) {
1267 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1268 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1269 else
1270 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1271 }
1272 return 1;
1273}
1274
ccd979bd
MF		 1275static int ocfs2_parse_options(struct super_block *sb,
1276 char *options,
c0123ade 1277 struct mount_options *mopt,
ccd979bd
MF		 1278 int is_remount)
1279{
52c303c5 1280 int status, user_stack = 0;
ccd979bd 1281 char *p;
2c442719 1282 u32 tmp;
ccd979bd 1283
32a42d39 1284 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
ccd979bd 1285
d147b3d6 1286 mopt->commit_interval = 0;
4b37fcb7 1287 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
c0123ade
TY		 1288 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1289 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1290 mopt->localalloc_opt = -1;
b61817e1 1291 mopt->cluster_stack[0] = '\0';
d02f00cc 1292 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
83f92318 1293 mopt->dir_resv_level = -1;
ccd979bd
MF		 1294
1295 if (!options) {
1296 status = 1;
1297 goto bail;
1298 }
1299
1300 while ((p = strsep(&options, ",")) != NULL) {
1301 int token, option;
1302 substring_t args[MAX_OPT_ARGS];
1303
1304 if (!*p)
1305 continue;
1306
1307 token = match_token(p, tokens, args);
1308 switch (token) {
1309 case Opt_hb_local:
c0123ade 1310 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1311 break;
1312 case Opt_hb_none:
2c442719
SM		 1313 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1314 break;
1315 case Opt_hb_global:
1316 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
ccd979bd
MF		 1317 break;
1318 case Opt_barrier:
1319 if (match_int(&args[0], &option)) {
1320 status = 0;
1321 goto bail;
1322 }
1323 if (option)
c0123ade 1324 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1325 else
c0123ade 1326 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1327 break;
1328 case Opt_intr:
c0123ade 1329 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1330 break;
1331 case Opt_nointr:
c0123ade 1332 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1333 break;
1334 case Opt_err_panic:
7d0fb914
GR		 1335 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1336 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
c0123ade 1337 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1338 break;
1339 case Opt_err_ro:
7d0fb914 1340 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
c0123ade 1341 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
7d0fb914
GR		 1342 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1343 break;
1344 case Opt_err_cont:
1345 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1346 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1347 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
ccd979bd
MF		 1348 break;
1349 case Opt_data_ordered:
c0123ade 1350 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1351 break;
1352 case Opt_data_writeback:
c0123ade 1353 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1354 break;
cf1d6c76
TY		 1355 case Opt_user_xattr:
1356 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1357 break;
1358 case Opt_nouser_xattr:
1359 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1360 break;
7f1a37e3
TY		 1361 case Opt_atime_quantum:
1362 if (match_int(&args[0], &option)) {
1363 status = 0;
1364 goto bail;
1365 }
1366 if (option >= 0)
c0123ade 1367 mopt->atime_quantum = option;
7f1a37e3 1368 break;
baf4661a
SM		 1369 case Opt_slot:
1370 option = 0;
1371 if (match_int(&args[0], &option)) {
1372 status = 0;
1373 goto bail;
1374 }
1375 if (option)
c0123ade 1376 mopt->slot = (s16)option;
baf4661a 1377 break;
d147b3d6
MF		 1378 case Opt_commit:
1379 option = 0;
1380 if (match_int(&args[0], &option)) {
1381 status = 0;
1382 goto bail;
1383 }
1384 if (option < 0)
1385 return 0;
1386 if (option == 0)
2b4e30fb 1387 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1388 mopt->commit_interval = HZ * option;
1389 break;
2fbe8d1e
SM		 1390 case Opt_localalloc:
1391 option = 0;
1392 if (match_int(&args[0], &option)) {
1393 status = 0;
1394 goto bail;
1395 }
73c8a800 1396 if (option >= 0)
2fbe8d1e
SM		 1397 mopt->localalloc_opt = option;
1398 break;
53fc622b
MF		 1399 case Opt_localflocks:
1400 /*
1401 * Changing this during remount could race
1402 * flock() requests, or "unbalance" existing
1403 * ones (e.g., a lock is taken in one mode but
1404 * dropped in the other). If users care enough
1405 * to flip locking modes during remount, we
1406 * could add a "local" flag to individual
1407 * flock structures for proper tracking of
1408 * state.
1409 */
1410 if (!is_remount)
1411 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1412 break;
b61817e1
JB		 1413 case Opt_stack:
1414 /* Check both that the option we were passed
1415 * is of the right length and that it is a proper
1416 * string of the right length.
1417 */
1418 if (((args[0].to - args[0].from) !=
1419 OCFS2_STACK_LABEL_LEN) ||
1420 (strnlen(args[0].from,
1421 OCFS2_STACK_LABEL_LEN) !=
1422 OCFS2_STACK_LABEL_LEN)) {
1423 mlog(ML_ERROR,
1424 "Invalid cluster_stack option\n");
1425 status = 0;
1426 goto bail;
1427 }
1428 memcpy(mopt->cluster_stack, args[0].from,
1429 OCFS2_STACK_LABEL_LEN);
1430 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
52c303c5
MF		 1431 /*
1432 * Open code the memcmp here as we don't have
1433 * an osb to pass to
1434 * ocfs2_userspace_stack().
1435 */
1436 if (memcmp(mopt->cluster_stack,
1437 OCFS2_CLASSIC_CLUSTER_STACK,
1438 OCFS2_STACK_LABEL_LEN))
1439 user_stack = 1;
b61817e1 1440 break;
12462f1d
JB		 1441 case Opt_inode64:
1442 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1443 break;
19ece546 1444 case Opt_usrquota:
19ece546
JK		 1445 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1446 break;
1447 case Opt_grpquota:
19ece546
JK		 1448 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1449 break;
7bdb0d18
TY		 1450 case Opt_coherency_buffered:
1451 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1452 break;
1453 case Opt_coherency_full:
1454 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1455 break;
a68979b8
TY		 1456 case Opt_acl:
1457 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1458 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1459 break;
1460 case Opt_noacl:
5297aad8 1461 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1462 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1463 break;
d02f00cc
MF		 1464 case Opt_resv_level:
1465 if (is_remount)
1466 break;
1467 if (match_int(&args[0], &option)) {
1468 status = 0;
1469 goto bail;
1470 }
1471 if (option >= OCFS2_MIN_RESV_LEVEL &&
1472 option < OCFS2_MAX_RESV_LEVEL)
1473 mopt->resv_level = option;
1474 break;
83f92318
MF		 1475 case Opt_dir_resv_level:
1476 if (is_remount)
1477 break;
1478 if (match_int(&args[0], &option)) {
1479 status = 0;
1480 goto bail;
1481 }
1482 if (option >= OCFS2_MIN_RESV_LEVEL &&
1483 option < OCFS2_MAX_RESV_LEVEL)
1484 mopt->dir_resv_level = option;
1485 break;
1dfeb768 1486 case Opt_journal_async_commit:
1487 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1488 break;
ccd979bd
MF		 1489 default:
1490 mlog(ML_ERROR,
1491 "Unrecognized mount option \"%s\" "
1492 "or missing value\n", p);
1493 status = 0;
1494 goto bail;
1495 }
1496 }
1497
52c303c5
MF		 1498 if (user_stack == 0) {
1499 /* Ensure only one heartbeat mode */
1500 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1501 OCFS2_MOUNT_HB_GLOBAL |
1502 OCFS2_MOUNT_HB_NONE);
1503 if (hweight32(tmp) != 1) {
1504 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1505 status = 0;
1506 goto bail;
1507 }
2c442719
SM		 1508 }
1509
ccd979bd
MF		 1510 status = 1;
1511
1512bail:
ccd979bd
MF		 1513 return status;
1514}
1515
34c80b1d 1516static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
d550071c 1517{
34c80b1d 1518 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
d550071c 1519 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1520 unsigned int local_alloc_megs;
d550071c 1521
2c442719
SM		 1522 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1523 seq_printf(s, ",_netdev");
1524 if (opts & OCFS2_MOUNT_HB_LOCAL)
1525 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1526 else
1527 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1528 } else
1529 seq_printf(s, ",%s", OCFS2_HB_NONE);
d550071c
SM		 1530
1531 if (opts & OCFS2_MOUNT_NOINTR)
1532 seq_printf(s, ",nointr");
1533
1534 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1535 seq_printf(s, ",data=writeback");
1536 else
1537 seq_printf(s, ",data=ordered");
1538
1539 if (opts & OCFS2_MOUNT_BARRIER)
1540 seq_printf(s, ",barrier=1");
1541
1542 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1543 seq_printf(s, ",errors=panic");
7d0fb914
GR		 1544 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1545 seq_printf(s, ",errors=continue");
d550071c
SM		 1546 else
1547 seq_printf(s, ",errors=remount-ro");
1548
1549 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1550 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1551
34c80b1d 1552 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
d550071c 1553
d147b3d6
MF		 1554 if (osb->osb_commit_interval)
1555 seq_printf(s, ",commit=%u",
1556 (unsigned) (osb->osb_commit_interval / HZ));
1557
ebcee4b5 1558 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
6b82021b 1559 if (local_alloc_megs != ocfs2_la_default_mb(osb))
ebcee4b5 1560 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1561
53fc622b
MF		 1562 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1563 seq_printf(s, ",localflocks,");
1564
b61817e1
JB		 1565 if (osb->osb_cluster_stack[0])
1566 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1567 osb->osb_cluster_stack);
19ece546
JK		 1568 if (opts & OCFS2_MOUNT_USRQUOTA)
1569 seq_printf(s, ",usrquota");
1570 if (opts & OCFS2_MOUNT_GRPQUOTA)
1571 seq_printf(s, ",grpquota");
b61817e1 1572
7bdb0d18
TY		 1573 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1574 seq_printf(s, ",coherency=buffered");
1575 else
1576 seq_printf(s, ",coherency=full");
1577
b0f73cfc
SM		 1578 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1579 seq_printf(s, ",nouser_xattr");
1580 else
1581 seq_printf(s, ",user_xattr");
1582
12462f1d
JB		 1583 if (opts & OCFS2_MOUNT_INODE64)
1584 seq_printf(s, ",inode64");
1585
a68979b8
TY		 1586 if (opts & OCFS2_MOUNT_POSIX_ACL)
1587 seq_printf(s, ",acl");
1588 else
1589 seq_printf(s, ",noacl");
a68979b8 1590
d02f00cc
MF		 1591 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1592 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1593
83f92318
MF		 1594 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1595 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1596
1dfeb768 1597 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1598 seq_printf(s, ",journal_async_commit");
1599
d550071c
SM		 1600 return 0;
1601}
1602
ccd979bd
MF		 1603static int __init ocfs2_init(void)
1604{
c18ceab0 1605 int status;
a11f7e63 1606
ccd979bd 1607 status = init_ocfs2_uptodate_cache();
342827d7
AV		 1608 if (status < 0)
1609 goto out1;
ccd979bd
MF		 1610
1611 status = ocfs2_initialize_mem_caches();
342827d7
AV		 1612 if (status < 0)
1613 goto out2;
ccd979bd
MF		 1614
1615 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1616 if (!ocfs2_wq) {
1617 status = -ENOMEM;
342827d7 1618 goto out3;
ccd979bd
MF		 1619 }
1620
ccd979bd 1621 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
8f443e23
LT		 1622 if (!ocfs2_debugfs_root) {
1623 status = -ENOMEM;
ccd979bd 1624 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
dc171580 1625 goto out4;
ccd979bd
MF		 1626 }
1627
63e0c48a
JB		 1628 ocfs2_set_locking_protocol();
1629
9e33d69f 1630 status = register_quota_format(&ocfs2_quota_format);
342827d7
AV		 1631 if (status < 0)
1632 goto out4;
1633 status = register_filesystem(&ocfs2_fs_type);
1634 if (!status)
1635 return 0;
ccd979bd 1636
342827d7
AV		 1637 unregister_quota_format(&ocfs2_quota_format);
1638out4:
1639 destroy_workqueue(ocfs2_wq);
1640 debugfs_remove(ocfs2_debugfs_root);
1641out3:
1642 ocfs2_free_mem_caches();
1643out2:
1644 exit_ocfs2_uptodate_cache();
1645out1:
1646 mlog_errno(status);
1647 return status;
ccd979bd
MF		 1648}
1649
1650static void __exit ocfs2_exit(void)
1651{
ccd979bd
MF		 1652 if (ocfs2_wq) {
1653 flush_workqueue(ocfs2_wq);
1654 destroy_workqueue(ocfs2_wq);
1655 }
1656
9e33d69f
JK		 1657 unregister_quota_format(&ocfs2_quota_format);
1658
ccd979bd
MF		 1659 debugfs_remove(ocfs2_debugfs_root);
1660
1661 ocfs2_free_mem_caches();
1662
1663 unregister_filesystem(&ocfs2_fs_type);
1664
ccd979bd 1665 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1666}
1667
1668static void ocfs2_put_super(struct super_block *sb)
1669{
32a42d39 1670 trace_ocfs2_put_super(sb);
ccd979bd
MF		 1671
1672 ocfs2_sync_blockdev(sb);
1673 ocfs2_dismount_volume(sb, 0);
ccd979bd
MF		 1674}
1675
726c3342 1676static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1677{
1678 struct ocfs2_super *osb;
1679 u32 numbits, freebits;
1680 int status;
1681 struct ocfs2_dinode *bm_lock;
1682 struct buffer_head *bh = NULL;
1683 struct inode *inode = NULL;
1684
32a42d39 1685 trace_ocfs2_statfs(dentry->d_sb, buf);
ccd979bd 1686
726c3342 1687 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1688
1689 inode = ocfs2_get_system_file_inode(osb,
1690 GLOBAL_BITMAP_SYSTEM_INODE,
1691 OCFS2_INVALID_SLOT);
1692 if (!inode) {
1693 mlog(ML_ERROR, "failed to get bitmap inode\n");
1694 status = -EIO;
1695 goto bail;
1696 }
1697
e63aecb6 1698 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1699 if (status < 0) {
1700 mlog_errno(status);
1701 goto bail;
1702 }
1703
1704 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1705
1706 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1707 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1708
1709 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1710 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1711 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1712 buf->f_blocks = ((sector_t) numbits) *
1713 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1714 buf->f_bfree = ((sector_t) freebits) *
1715 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1716 buf->f_bavail = buf->f_bfree;
1717 buf->f_files = numbits;
1718 buf->f_ffree = freebits;
837711f8
CL		 1719 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1720 & 0xFFFFFFFFUL;
1721 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1722 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF		 1723
1724 brelse(bh);
1725
e63aecb6 1726 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1727 status = 0;
1728bail:
1729 if (inode)
1730 iput(inode);
1731
c1e8d35e
TM		 1732 if (status)
1733 mlog_errno(status);
ccd979bd
MF		 1734
1735 return status;
1736}
1737
51cc5068 1738static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1739{
1740 struct ocfs2_inode_info *oi = data;
1741
a35afb83
CL		 1742 oi->ip_flags = 0;
1743 oi->ip_open_count = 0;
1744 spin_lock_init(&oi->ip_lock);
1745 ocfs2_extent_map_init(&oi->vfs_inode);
1746 INIT_LIST_HEAD(&oi->ip_io_markers);
a35afb83 1747 oi->ip_dir_start_lookup = 0;
c18ceab0 1748 mutex_init(&oi->ip_unaligned_aio);
a35afb83 1749 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1750 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1751 mutex_init(&oi->ip_io_mutex);
ccd979bd 1752
a35afb83
CL		 1753 oi->ip_blkno = 0ULL;
1754 oi->ip_clusters = 0;
ccd979bd 1755
4fe370af
MF		 1756 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1757
a35afb83 1758 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1759 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1760 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1761
8cb471e8 1762 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1763 &ocfs2_inode_caching_ops);
ccd979bd 1764
a35afb83 1765 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1766}
1767
1768static int ocfs2_initialize_mem_caches(void)
1769{
1770 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1771 sizeof(struct ocfs2_inode_info),
1772 0,
1773 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1774 SLAB_MEM_SPREAD),
20c2df83 1775 ocfs2_inode_init_once);
9e33d69f
JK		 1776 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1777 sizeof(struct ocfs2_dquot),
1778 0,
1779 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1780 SLAB_MEM_SPREAD),
1781 NULL);
1782 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1783 sizeof(struct ocfs2_quota_chunk),
1784 0,
1785 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1786 NULL);
1787 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1788 !ocfs2_qf_chunk_cachep) {
1789 if (ocfs2_inode_cachep)
1790 kmem_cache_destroy(ocfs2_inode_cachep);
1791 if (ocfs2_dquot_cachep)
1792 kmem_cache_destroy(ocfs2_dquot_cachep);
1793 if (ocfs2_qf_chunk_cachep)
1794 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1795 return -ENOMEM;
9e33d69f 1796 }
ccd979bd 1797
ccd979bd
MF		 1798 return 0;
1799}
1800
1801static void ocfs2_free_mem_caches(void)
1802{
8c0a8537
KS		 1803 /*
1804 * Make sure all delayed rcu free inodes are flushed before we
1805 * destroy cache.
1806 */
1807 rcu_barrier();
ccd979bd
MF		 1808 if (ocfs2_inode_cachep)
1809 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1810 ocfs2_inode_cachep = NULL;
9e33d69f
JK		 1811
1812 if (ocfs2_dquot_cachep)
1813 kmem_cache_destroy(ocfs2_dquot_cachep);
1814 ocfs2_dquot_cachep = NULL;
1815
1816 if (ocfs2_qf_chunk_cachep)
1817 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1818 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1819}
1820
1821static int ocfs2_get_sector(struct super_block *sb,
1822 struct buffer_head **bh,
1823 int block,
1824 int sect_size)
1825{
1826 if (!sb_set_blocksize(sb, sect_size)) {
1827 mlog(ML_ERROR, "unable to set blocksize\n");
1828 return -EIO;
1829 }
1830
1831 *bh = sb_getblk(sb, block);
1832 if (!*bh) {
7391a294
RX		 1833 mlog_errno(-ENOMEM);
1834 return -ENOMEM;
ccd979bd
MF		 1835 }
1836 lock_buffer(*bh);
1837 if (!buffer_dirty(*bh))
1838 clear_buffer_uptodate(*bh);
1839 unlock_buffer(*bh);
1840 ll_rw_block(READ, 1, bh);
1841 wait_on_buffer(*bh);
28d57d43 1842 if (!buffer_uptodate(*bh)) {
1843 mlog_errno(-EIO);
1844 brelse(*bh);
1845 *bh = NULL;
1846 return -EIO;
1847 }
1848
ccd979bd
MF		 1849 return 0;
1850}
1851
ccd979bd
MF		 1852static int ocfs2_mount_volume(struct super_block *sb)
1853{
1854 int status = 0;
1855 int unlock_super = 0;
1856 struct ocfs2_super *osb = OCFS2_SB(sb);
1857
ccd979bd
MF		 1858 if (ocfs2_is_hard_readonly(osb))
1859 goto leave;
1860
ccd979bd
MF		 1861 status = ocfs2_dlm_init(osb);
1862 if (status < 0) {
1863 mlog_errno(status);
1864 goto leave;
1865 }
1866
ccd979bd
MF		 1867 status = ocfs2_super_lock(osb, 1);
1868 if (status < 0) {
1869 mlog_errno(status);
1870 goto leave;
1871 }
1872 unlock_super = 1;
1873
1874 /* This will load up the node map and add ourselves to it. */
1875 status = ocfs2_find_slot(osb);
1876 if (status < 0) {
1877 mlog_errno(status);
1878 goto leave;
1879 }
1880
ccd979bd
MF		 1881 /* load all node-local system inodes */
1882 status = ocfs2_init_local_system_inodes(osb);
1883 if (status < 0) {
1884 mlog_errno(status);
1885 goto leave;
1886 }
1887
1888 status = ocfs2_check_volume(osb);
1889 if (status < 0) {
1890 mlog_errno(status);
1891 goto leave;
1892 }
1893
1894 status = ocfs2_truncate_log_init(osb);
06c59bb8 1895 if (status < 0)
ccd979bd 1896 mlog_errno(status);
c271c5c2 1897
ccd979bd
MF		 1898leave:
1899 if (unlock_super)
1900 ocfs2_super_unlock(osb, 1);
1901
ccd979bd
MF		 1902 return status;
1903}
1904
ccd979bd
MF		 1905static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1906{
286eaa95 1907 int tmp, hangup_needed = 0;
ccd979bd 1908 struct ocfs2_super *osb = NULL;
99d7a882 1909 char nodestr[12];
ccd979bd 1910
32a42d39 1911 trace_ocfs2_dismount_volume(sb);
ccd979bd
MF		 1912
1913 BUG_ON(!sb);
1914 osb = OCFS2_SB(sb);
1915 BUG_ON(!osb);
1916
50397507
SM		 1917 debugfs_remove(osb->osb_ctxt);
1918
692684e1
SM		 1919 /* Orphan scan should be stopped as early as possible */
1920 ocfs2_orphan_scan_stop(osb);
1921
19ece546
JK		 1922 ocfs2_disable_quotas(osb);
1923
e3a767b6
JK		 1924 /* All dquots should be freed by now */
1925 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1926 /* Wait for worker to be done with the work structure in osb */
1927 cancel_work_sync(&osb->dquot_drop_work);
1928
ccd979bd
MF		 1929 ocfs2_shutdown_local_alloc(osb);
1930
b253bfd8
X		 1931 ocfs2_truncate_log_shutdown(osb);
1932
553abd04
JB		 1933 /* This will disable recovery and flush any recovery work. */
1934 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1935
1936 ocfs2_journal_shutdown(osb);
1937
1938 ocfs2_sync_blockdev(sb);
1939
374a263e
TM		 1940 ocfs2_purge_refcount_trees(osb);
1941
4670c46d
JB		 1942 /* No cluster connection means we've failed during mount, so skip
1943 * all the steps which depended on that to complete. */
1944 if (osb->cconn) {
ccd979bd
MF		 1945 tmp = ocfs2_super_lock(osb, 1);
1946 if (tmp < 0) {
1947 mlog_errno(tmp);
1948 return;
1949 }
19b613d4 1950 }
ccd979bd 1951
19b613d4
MF		 1952 if (osb->slot_num != OCFS2_INVALID_SLOT)
1953 ocfs2_put_slot(osb);
ccd979bd 1954
4670c46d 1955 if (osb->cconn)
ccd979bd 1956 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1957
1958 ocfs2_release_system_inodes(osb);
1959
6953b4c0 1960 /*
6953b4c0
JB		 1961 * If we're dismounting due to mount error, mount.ocfs2 will clean
1962 * up heartbeat. If we're a local mount, there is no heartbeat.
1963 * If we failed before we got a uuid_str yet, we can't stop
1964 * heartbeat. Otherwise, do it.
1965 */
03efed8a
TY		 1966 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1967 !ocfs2_is_hard_readonly(osb))
286eaa95
JB		 1968 hangup_needed = 1;
1969
1970 if (osb->cconn)
1971 ocfs2_dlm_shutdown(osb, hangup_needed);
1972
73be192b 1973 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1974 debugfs_remove(osb->osb_debug_root);
1975
1976 if (hangup_needed)
6953b4c0 1977 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1978
1979 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1980
c271c5c2
SM		 1981 if (ocfs2_mount_local(osb))
1982 snprintf(nodestr, sizeof(nodestr), "local");
1983 else
19fdb624 1984 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1985
1986 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1987 osb->dev_str, nodestr);
ccd979bd
MF		 1988
1989 ocfs2_delete_osb(osb);
1990 kfree(osb);
1991 sb->s_dev = 0;
1992 sb->s_fs_info = NULL;
1993}
1994
1995static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1996 unsigned uuid_bytes)
1997{
1998 int i, ret;
1999 char *ptr;
2000
2001 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2002
cd861280 2003 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 2004 if (osb->uuid_str == NULL)
2005 return -ENOMEM;
2006
ccd979bd
MF		 2007 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2008 /* print with null */
2009 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2010 if (ret != 2) /* drop super cleans up */
2011 return -EINVAL;
2012 /* then only advance past the last char */
2013 ptr += 2;
2014 }
2015
2016 return 0;
2017}
2018
3bdb8efd
PL		 2019/* Make sure entire volume is addressable by our journal. Requires
2020 osb_clusters_at_boot to be valid and for the journal to have been
2021 initialized by ocfs2_journal_init(). */
2022static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2023{
2024 int status = 0;
2025 u64 max_block =
2026 ocfs2_clusters_to_blocks(osb->sb,
2027 osb->osb_clusters_at_boot) - 1;
2028
2029 /* 32-bit block number is always OK. */
2030 if (max_block <= (u32)~0ULL)
2031 goto out;
2032
2033 /* Volume is "huge", so see if our journal is new enough to
2034 support it. */
2035 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2036 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2037 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2038 JBD2_FEATURE_INCOMPAT_64BIT))) {
2039 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2040 "Enable the 'block64' journal option with tunefs.ocfs2");
2041 status = -EFBIG;
2042 goto out;
2043 }
2044
2045 out:
2046 return status;
2047}
2048
ccd979bd
MF		 2049static int ocfs2_initialize_super(struct super_block *sb,
2050 struct buffer_head *bh,
73be192b
JB		 2051 int sector_size,
2052 struct ocfs2_blockcheck_stats *stats)
ccd979bd 2053{
bddb8eb3 2054 int status;
5a254031
MF		 2055 int i, cbits, bbits;
2056 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 2057 struct inode *inode = NULL;
ccd979bd 2058 struct ocfs2_journal *journal;
ccd979bd 2059 struct ocfs2_super *osb;
3bdb8efd 2060 u64 total_blocks;
ccd979bd 2061
cd861280 2062 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 2063 if (!osb) {
2064 status = -ENOMEM;
2065 mlog_errno(status);
2066 goto bail;
2067 }
2068
2069 sb->s_fs_info = osb;
2070 sb->s_op = &ocfs2_sops;
ba87167c 2071 sb->s_d_op = &ocfs2_dentry_ops;
ccd979bd 2072 sb->s_export_op = &ocfs2_export_ops;
664dbd5f 2073 sb->s_qcop = &dquot_quotactl_sysfile_ops;
19ece546 2074 sb->dq_op = &ocfs2_quota_operations;
1c92ec67 2075 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cf1d6c76 2076 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 2077 sb->s_time_gran = 1;
ccd979bd
MF		 2078 sb->s_flags |= MS_NOATIME;
2079 /* this is needed to support O_LARGEFILE */
5a254031
MF		 2080 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2081 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2082 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
58be19dc
VD		 2083 memcpy(sb->s_uuid, di->id2.i_super.s_uuid,
2084 sizeof(di->id2.i_super.s_uuid));
ccd979bd 2085
9b7895ef
MF		 2086 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2087
2088 for (i = 0; i < 3; i++)
2089 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2090 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2091
ccd979bd
MF		 2092 osb->sb = sb;
2093 /* Save off for ocfs2_rw_direct */
2094 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2095 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2096
34d024f8
MF		 2097 spin_lock_init(&osb->dc_task_lock);
2098 init_waitqueue_head(&osb->dc_event);
2099 osb->dc_work_sequence = 0;
2100 osb->dc_wake_sequence = 0;
ccd979bd
MF		 2101 INIT_LIST_HEAD(&osb->blocked_lock_list);
2102 osb->blocked_lock_count = 0;
ccd979bd 2103 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2104 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2105 ocfs2_init_steal_slots(osb);
ccd979bd 2106
43b10a20 2107 mutex_init(&osb->system_file_mutex);
2108
ccd979bd
MF		 2109 atomic_set(&osb->alloc_stats.moves, 0);
2110 atomic_set(&osb->alloc_stats.local_data, 0);
2111 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2112 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2113 atomic_set(&osb->alloc_stats.bg_extends, 0);
2114
73be192b
JB		 2115 /* Copy the blockcheck stats from the superblock probe */
2116 osb->osb_ecc_stats = *stats;
2117
ccd979bd
MF		 2118 ocfs2_init_node_maps(osb);
2119
2120 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2121 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2122
75d9bbc7
GR		 2123 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2124 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2125 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2126 osb->max_slots);
2127 status = -EINVAL;
2128 goto bail;
2129 }
75d9bbc7 2130
8b712cd5
JM		 2131 ocfs2_orphan_scan_init(osb);
2132
553abd04
JB		 2133 status = ocfs2_recovery_init(osb);
2134 if (status) {
2135 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2136 mlog_errno(status);
2137 goto bail;
2138 }
ccd979bd
MF		 2139
2140 init_waitqueue_head(&osb->checkpoint_event);
ccd979bd 2141
7f1a37e3
TY		 2142 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2143
ccd979bd
MF		 2144 osb->slot_num = OCFS2_INVALID_SLOT;
2145
8154da3d
TY		 2146 osb->s_xattr_inline_size = le16_to_cpu(
2147 di->id2.i_super.s_xattr_inline_size);
fdd77704 2148
ccd979bd
MF		 2149 osb->local_alloc_state = OCFS2_LA_UNUSED;
2150 osb->local_alloc_bh = NULL;
9c7af40b 2151 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2152
ccd979bd
MF		 2153 init_waitqueue_head(&osb->osb_mount_event);
2154
d02f00cc
MF		 2155 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2156 if (status) {
2157 mlog_errno(status);
2158 goto bail;
2159 }
2160
ccd979bd
MF		 2161 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2162 if (!osb->vol_label) {
2163 mlog(ML_ERROR, "unable to alloc vol label\n");
2164 status = -ENOMEM;
2165 goto bail;
2166 }
2167
539d8264
SM		 2168 osb->slot_recovery_generations =
2169 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2170 GFP_KERNEL);
2171 if (!osb->slot_recovery_generations) {
2172 status = -ENOMEM;
2173 mlog_errno(status);
2174 goto bail;
2175 }
2176
b4df6ed8
MF		 2177 init_waitqueue_head(&osb->osb_wipe_event);
2178 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2179 sizeof(*osb->osb_orphan_wipes),
2180 GFP_KERNEL);
2181 if (!osb->osb_orphan_wipes) {
2182 status = -ENOMEM;
2183 mlog_errno(status);
2184 goto bail;
2185 }
2186
374a263e
TM		 2187 osb->osb_rf_lock_tree = RB_ROOT;
2188
ccd979bd
MF		 2189 osb->s_feature_compat =
2190 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2191 osb->s_feature_ro_compat =
2192 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2193 osb->s_feature_incompat =
2194 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2195
2196 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2197 mlog(ML_ERROR, "couldn't mount because of unsupported "
2198 "optional features (%x).\n", i);
2199 status = -EINVAL;
2200 goto bail;
2201 }
2202 if (!(osb->sb->s_flags & MS_RDONLY) &&
2203 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2204 mlog(ML_ERROR, "couldn't mount RDWR because of "
2205 "unsupported optional features (%x).\n", i);
2206 status = -EINVAL;
2207 goto bail;
2208 }
2209
98f486f2
SM		 2210 if (ocfs2_clusterinfo_valid(osb)) {
2211 osb->osb_stackflags =
2212 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
c74a3bdd 2213 strlcpy(osb->osb_cluster_stack,
b61817e1 2214 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
c74a3bdd 2215 OCFS2_STACK_LABEL_LEN + 1);
b61817e1
JB		 2216 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2217 mlog(ML_ERROR,
2218 "couldn't mount because of an invalid "
2219 "cluster stack label (%s) \n",
2220 osb->osb_cluster_stack);
2221 status = -EINVAL;
2222 goto bail;
2223 }
c74a3bdd
GR		 2224 strlcpy(osb->osb_cluster_name,
2225 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2226 OCFS2_CLUSTER_NAME_LEN + 1);
b61817e1
JB		 2227 } else {
2228 /* The empty string is identical with classic tools that
2229 * don't know about s_cluster_info. */
2230 osb->osb_cluster_stack[0] = '\0';
2231 }
2232
ccd979bd
MF		 2233 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2234
2235 /* FIXME
2236 * This should be done in ocfs2_journal_init(), but unknown
2237 * ordering issues will cause the filesystem to crash.
2238 * If anyone wants to figure out what part of the code
2239 * refers to osb->journal before ocfs2_journal_init() is run,
2240 * be my guest.
2241 */
2242 /* initialize our journal structure */
2243
cd861280 2244 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2245 if (!journal) {
2246 mlog(ML_ERROR, "unable to alloc journal\n");
2247 status = -ENOMEM;
2248 goto bail;
2249 }
2250 osb->journal = journal;
2251 journal->j_osb = osb;
2252
2253 atomic_set(&journal->j_num_trans, 0);
2254 init_rwsem(&journal->j_trans_barrier);
2255 init_waitqueue_head(&journal->j_checkpointed);
2256 spin_lock_init(&journal->j_lock);
2257 journal->j_trans_id = (unsigned long) 1;
2258 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2259 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2260 journal->j_state = OCFS2_JOURNAL_FREE;
2261
e3a767b6
JK		 2262 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2263 init_llist_head(&osb->dquot_drop_list);
2264
ccd979bd
MF		 2265 /* get some pseudo constants for clustersize bits */
2266 osb->s_clustersize_bits =
2267 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2268 osb->s_clustersize = 1 << osb->s_clustersize_bits;
ccd979bd
MF		 2269
2270 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2271 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2272 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2273 osb->s_clustersize);
2274 status = -EINVAL;
2275 goto bail;
2276 }
2277
3bdb8efd
PL		 2278 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2279 le32_to_cpu(di->i_clusters));
2280
2281 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2282 total_blocks);
2283 if (status) {
2284 mlog(ML_ERROR, "Volume too large "
2285 "to mount safely on this system");
2286 status = -EFBIG;
ccd979bd
MF		 2287 goto bail;
2288 }
2289
2290 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2291 sizeof(di->id2.i_super.s_uuid))) {
2292 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2293 status = -ENOMEM;
2294 goto bail;
2295 }
2296
69201bb1
FF		 2297 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2298 OCFS2_MAX_VOL_LABEL_LEN);
ccd979bd
MF		 2299 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2300 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2301 osb->first_cluster_group_blkno =
2302 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2303 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2304 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
32a42d39
TM		 2305 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2306 (unsigned long long)osb->root_blkno,
2307 (unsigned long long)osb->system_dir_blkno,
2308 osb->s_clustersize_bits);
ccd979bd
MF		 2309
2310 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2311 if (!osb->osb_dlm_debug) {
2312 status = -ENOMEM;
2313 mlog_errno(status);
2314 goto bail;
2315 }
2316
2317 atomic_set(&osb->vol_state, VOLUME_INIT);
2318
2319 /* load root, system_dir, and all global system inodes */
2320 status = ocfs2_init_global_system_inodes(osb);
2321 if (status < 0) {
2322 mlog_errno(status);
2323 goto bail;
2324 }
2325
2326 /*
2327 * global bitmap
2328 */
2329 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2330 OCFS2_INVALID_SLOT);
2331 if (!inode) {
2332 status = -EINVAL;
2333 mlog_errno(status);
2334 goto bail;
2335 }
2336
2337 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
6b82021b 2338 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
ccd979bd 2339 iput(inode);
ccd979bd 2340
8571882c
TM		 2341 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2342 osb->s_feature_incompat) * 8;
ccd979bd
MF		 2343
2344 status = ocfs2_init_slot_info(osb);
2345 if (status < 0) {
2346 mlog_errno(status);
2347 goto bail;
2348 }
9de16262 2349 cleancache_init_shared_fs(sb);
ccd979bd 2350
ccd979bd 2351bail:
ccd979bd
MF		 2352 return status;
2353}
2354
2355/*
2356 * will return: -EAGAIN if it is ok to keep searching for superblocks
2357 * -EINVAL if there is a bad superblock
2358 * 0 on success
2359 */
2360static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2361 struct buffer_head *bh,
73be192b
JB		 2362 u32 blksz,
2363 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2364{
2365 int status = -EAGAIN;
2366
ccd979bd
MF		 2367 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2368 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2369 /* We have to do a raw check of the feature here */
2370 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2371 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2372 status = ocfs2_block_check_validate(bh->b_data,
2373 bh->b_size,
73be192b
JB		 2374 &di->i_check,
2375 stats);
d030cc97
JB		 2376 if (status)
2377 goto out;
2378 }
ccd979bd
MF		 2379 status = -EINVAL;
2380 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2381 mlog(ML_ERROR, "found superblock with incorrect block "
2382 "size: found %u, should be %u\n",
2383 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2384 blksz);
2385 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2386 OCFS2_MAJOR_REV_LEVEL ||
2387 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2388 OCFS2_MINOR_REV_LEVEL) {
2389 mlog(ML_ERROR, "found superblock with bad version: "
2390 "found %u.%u, should be %u.%u\n",
2391 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2392 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2393 OCFS2_MAJOR_REV_LEVEL,
2394 OCFS2_MINOR_REV_LEVEL);
2395 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2396 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2397 "found %llu, should be %llu\n",
1ca1a111 2398 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2399 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2400 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2401 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2402 mlog(ML_ERROR, "bad cluster size found: %u\n",
2403 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2404 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2405 mlog(ML_ERROR, "bad root_blkno: 0\n");
2406 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2407 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2408 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2409 mlog(ML_ERROR,
2410 "Superblock slots found greater than file system "
2411 "maximum: found %u, max %u\n",
2412 le16_to_cpu(di->id2.i_super.s_max_slots),
2413 OCFS2_MAX_SLOTS);
2414 } else {
2415 /* found it! */
2416 status = 0;
2417 }
2418 }
2419
d030cc97 2420out:
c1e8d35e
TM		 2421 if (status && status != -EAGAIN)
2422 mlog_errno(status);
ccd979bd
MF		 2423 return status;
2424}
2425
2426static int ocfs2_check_volume(struct ocfs2_super *osb)
2427{
bddb8eb3 2428 int status;
ccd979bd 2429 int dirty;
c271c5c2 2430 int local;
ccd979bd
MF		 2431 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2432 * recover
2433 * ourselves. */
2434
ccd979bd
MF		 2435 /* Init our journal object. */
2436 status = ocfs2_journal_init(osb->journal, &dirty);
2437 if (status < 0) {
2438 mlog(ML_ERROR, "Could not initialize journal!\n");
2439 goto finally;
2440 }
2441
3bdb8efd
PL		 2442 /* Now that journal has been initialized, check to make sure
2443 entire volume is addressable. */
2444 status = ocfs2_journal_addressable(osb);
2445 if (status)
2446 goto finally;
2447
ccd979bd
MF		 2448 /* If the journal was unmounted cleanly then we don't want to
2449 * recover anything. Otherwise, journal_load will do that
2450 * dirty work for us :) */
2451 if (!dirty) {
2452 status = ocfs2_journal_wipe(osb->journal, 0);
2453 if (status < 0) {
2454 mlog_errno(status);
2455 goto finally;
2456 }
2457 } else {
619c200d
SM		 2458 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2459 "unmounted cleanly, recovering it.\n", osb->dev_str);
ccd979bd
MF		 2460 }
2461
c271c5c2
SM		 2462 local = ocfs2_mount_local(osb);
2463
ccd979bd 2464 /* will play back anything left in the journal. */
539d8264 2465 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2466 if (status < 0) {
2467 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2468 goto finally;
2469 }
ccd979bd 2470
1dfeb768 2471 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2472 jbd2_journal_set_features(osb->journal->j_journal,
2473 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2474 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2475 else
2476 jbd2_journal_clear_features(osb->journal->j_journal,
2477 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2478 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2479
ccd979bd
MF		 2480 if (dirty) {
2481 /* recover my local alloc if we didn't unmount cleanly. */
2482 status = ocfs2_begin_local_alloc_recovery(osb,
2483 osb->slot_num,
2484 &local_alloc);
2485 if (status < 0) {
2486 mlog_errno(status);
2487 goto finally;
2488 }
2489 /* we complete the recovery process after we've marked
2490 * ourselves as mounted. */
2491 }
2492
ccd979bd
MF		 2493 status = ocfs2_load_local_alloc(osb);
2494 if (status < 0) {
2495 mlog_errno(status);
2496 goto finally;
2497 }
2498
2499 if (dirty) {
2500 /* Recovery will be completed after we've mounted the
2501 * rest of the volume. */
2502 osb->dirty = 1;
2503 osb->local_alloc_copy = local_alloc;
2504 local_alloc = NULL;
2505 }
2506
2507 /* go through each journal, trylock it and if you get the
2508 * lock, and it's marked as dirty, set the bit in the recover
2509 * map and launch a recovery thread for it. */
2510 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2511 if (status < 0) {
2512 mlog_errno(status);
2513 goto finally;
2514 }
2515
2516 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2517 if (status < 0)
2518 mlog_errno(status);
2519
2520finally:
d787ab09 2521 kfree(local_alloc);
ccd979bd 2522
c1e8d35e
TM		 2523 if (status)
2524 mlog_errno(status);
ccd979bd
MF		 2525 return status;
2526}
2527
2528/*
2529 * The routine gets called from dismount or close whenever a dismount on
2530 * volume is requested and the osb open count becomes 1.
2531 * It will remove the osb from the global list and also free up all the
2532 * initialized resources and fileobject.
2533 */
2534static void ocfs2_delete_osb(struct ocfs2_super *osb)
2535{
ccd979bd
MF		 2536 /* This function assumes that the caller has the main osb resource */
2537
8e8a4603 2538 ocfs2_free_slot_info(osb);
ccd979bd 2539
b4df6ed8 2540 kfree(osb->osb_orphan_wipes);
539d8264 2541 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2542 /* FIXME
2543 * This belongs in journal shutdown, but because we have to
421f91d2 2544 * allocate osb->journal at the start of ocfs2_initialize_osb(),
ccd979bd
MF		 2545 * we free it here.
2546 */
2547 kfree(osb->journal);
d787ab09 2548 kfree(osb->local_alloc_copy);
ccd979bd 2549 kfree(osb->uuid_str);
f13a568e 2550 kfree(osb->vol_label);
ccd979bd
MF		 2551 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2552 memset(osb, 0, sizeof(struct ocfs2_super));
ccd979bd
MF		 2553}
2554
7d0fb914
GR		 2555/* Depending on the mount option passed, perform one of the following:
2556 * Put OCFS2 into a readonly state (default)
2557 * Return EIO so that only the process errs
2558 * Fix the error as if fsck.ocfs2 -y
2559 * panic
2560 */
2561static int ocfs2_handle_error(struct super_block *sb)
ccd979bd
MF		 2562{
2563 struct ocfs2_super *osb = OCFS2_SB(sb);
7d0fb914 2564 int rv = 0;
ccd979bd
MF		 2565
2566 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
7d0fb914
GR		 2567 pr_crit("On-disk corruption discovered. "
2568 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
ccd979bd 2569
7d0fb914
GR		 2570 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2571 panic("OCFS2: (device %s): panic forced after error\n",
2572 sb->s_id);
2573 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2574 pr_crit("OCFS2: Returning error to the calling process.\n");
2575 rv = -EIO;
2576 } else { /* default option */
2577 rv = -EROFS;
2578 if (sb->s_flags & MS_RDONLY &&
2579 (ocfs2_is_soft_readonly(osb) ||
2580 ocfs2_is_hard_readonly(osb)))
2581 return rv;
2582
2583 pr_crit("OCFS2: File system is now read-only.\n");
2584 sb->s_flags |= MS_RDONLY;
2585 ocfs2_set_ro_flag(osb, 0);
2586 }
2587
2588 return rv;
ccd979bd
MF		 2589}
2590
7d0fb914 2591int __ocfs2_error(struct super_block *sb, const char *function,
1543306e 2592 const char *fmt, ...)
ccd979bd 2593{
1543306e 2594 struct va_format vaf;
ccd979bd
MF		 2595 va_list args;
2596
2597 va_start(args, fmt);
1543306e
JP		 2598 vaf.fmt = fmt;
2599 vaf.va = &args;
ccd979bd
MF		 2600
2601 /* Not using mlog here because we want to show the actual
2602 * function the error came from. */
1543306e
JP		 2603 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV\n",
2604 sb->s_id, function, &vaf);
2605
2606 va_end(args);
ccd979bd 2607
7d0fb914 2608 return ocfs2_handle_error(sb);
ccd979bd
MF		 2609}
2610
2611/* Handle critical errors. This is intentionally more drastic than
2612 * ocfs2_handle_error, so we only use for things like journal errors,
2613 * etc. */
1543306e 2614void __ocfs2_abort(struct super_block *sb, const char *function,
ccd979bd
MF		 2615 const char *fmt, ...)
2616{
1543306e 2617 struct va_format vaf;
ccd979bd
MF		 2618 va_list args;
2619
2620 va_start(args, fmt);
ccd979bd 2621
1543306e
JP		 2622 vaf.fmt = fmt;
2623 vaf.va = &args;
2624
2625 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV\n",
2626 sb->s_id, function, &vaf);
2627
2628 va_end(args);
ccd979bd
MF		 2629
2630 /* We don't have the cluster support yet to go straight to
2631 * hard readonly in here. Until then, we want to keep
2632 * ocfs2_abort() so that we can at least mark critical
2633 * errors.
2634 *
2635 * TODO: This should abort the journal and alert other nodes
2636 * that our slot needs recovery. */
2637
2638 /* Force a panic(). This stinks, but it's better than letting
2639 * things continue without having a proper hard readonly
2640 * here. */
a2f2ddbf
SM		 2641 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2642 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 2643 ocfs2_handle_error(sb);
2644}
2645
e4b963f1
JB		 2646/*
2647 * Void signal blockers, because in-kernel sigprocmask() only fails
2648 * when SIG_* is wrong.
2649 */
2650void ocfs2_block_signals(sigset_t *oldset)
2651{
2652 int rc;
2653 sigset_t blocked;
2654
2655 sigfillset(&blocked);
2656 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2657 BUG_ON(rc);
2658}
2659
2660void ocfs2_unblock_signals(sigset_t *oldset)
2661{
2662 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2663 BUG_ON(rc);
2664}
2665
ccd979bd
MF		 2666module_init(ocfs2_init);
2667module_exit(ocfs2_exit);
Linux kernel - Deliverables
This page took 0.848246 seconds and 5 git commands to generate.